SHORT-TERM PREDICTION OF WIND SPEED BASED ON HYBRID DECOMPOSITION AND PCG-BiLSTM

In order to reduce the impact of the randomness of wind speed on wind power generation and improve the accuracy of wind speed short-term prediction. In this paper，a short-term wind speed prediction method based on hybrid decomposition，two-channel input and multi-branch PCG-BiLSTM deep learning model is proposed. Firstly，the annual wind speed data were divided into four seasons：spring，summer，autumn and winter，and spring was selected as the main experimental subjects. Secondly，singular spectrum decomposition（SSD）and variational modal decomposition（VMD）are used to reduce the complexity of the original spring wind speed data，generate subcomponents with different modalities and low complexity，and the two different modal subcomponents are combined into a mixed component to achieve the complementarity of the advantages of different mode decomposition algorithms. Finally，the mixed components are input to the multi-branch PCG-BiLSTM deep learning model in the form of two channels，and each branch of the model consists of a parallel convolutional neural network（CNN）and a gated recurrent unit（GRU）to obtained a spatio-temporal feature extraction module for extracting the spatio- temporal features of the mixed components of the combination of two decomposition components，and the spatio-temporal features of the corresponding mixed components extracted by each branch are aggregated and then further extracted by a bidirectional long short-term memory network（BiLSTM）for the forward and reverse bi-directional fluctuation regularity of the wind speed signals，and then the final wind speed prediction results are obtained. Several sets of experimental results show that the combined prediction method proposed in this paper has high accuracy and generalization ability in short-term wind speed prediction，which is better than other traditional prediction methods. © 2024 Science Press. All rights reserved.